Title: Taking zero lower bound seriously: A structural vector autoregression containing positive-valued components
Authors: Henri Nyberg - University of Turku and University of Helsinki (Finland) [presenting]
Abstract: In the conventional structural vector autoregressive (SVAR) analysis, the variables are treated as real-valued components. This is not, however, always the case as seen, for example, during the recent Zero Lower Bound state in nominal short-term interest rates. A SVAR model containing at least one strictly positive component is developed. Despite the nonlinearity originating from the positive-valued components, impulse response functions and forecast error variance decompositions can be expressed with explicit analytical formulae. The empirical results of the core three-variable U.S. monetary policy system generally point out important differences to the linear SVAR models.